Abstract: Abstract: In the past 20 years, protected agriculture has been developed rapidly in China, with a total area of 4,000 kha. A large number of mobile platforms are urgently needed to satisfy the need of operation, and research on autonomous vehicles with navigation ability has been highly valued. For structured hard pressed road surface between cultivation beds in modern greenhouse, curb-following navigation technology is most valuable for autonomous vehicle working in greenhouse. However, present curb-following navigation technologies based on machine vision are too complicated and lack of stability, while navigation technologies based on distance sensing cannot reach ideal accuracy of position/orientation detection. The target of curb-following navigation in this paper was to control the vehicle body within a stable position/orientation relative to the curb based on limited simple high/low level signals from arc array of photoelectric switches. Principle of detection and control of position/orientation based on both arc array of photoelectric switches and ideal control area was firstly introduced, and index Nd that represents the number of triggered photoelectric switches and index Nf which represents the center sequence number of triggered photoelectric switches were introduced to establish a position/orientation detection model. The position/orientation was classified by different thresholds of the two indices which can trigger the corresponding control program of trajectory. And then accurate position and orientation which were calculated by the values of two indexes were used to set the parameters of control program to realize the curb line following navigation in real time. This method can realize precise navigation with just limited number of high-low signals from photoelectric switches. The curb-following navigation accuracy relies on number of photoelectric switches, radius and central angle of the arc array, while triggered time interval between adjacent photoelectric switches must satisfy the need of response of sensing, control and mechanical transmission for a vehicle. Differential controlling strategies for different position/orientation states of vehicle based on arc array of photoelectric switches were also put forward. It was found from experiments that as the speed increased, both the transverse error and course angle of the vehicle displacement would rise. While within the designed speed, neither violent shock nor instability was found. Experiment results also indicated that the deviations of vehicle's position and orientation were kept -35 mm to +15 mm and -5° to +5°, respectively. Relative to the size of the greenhouse vehicle, the error was acceptable which satisfied the curb-following navigation requirement of pesticide spraying, seedling transplanting, transporting etc. in greenhouse. And the control cycle of trajectory was about 2 m along the curb line, which indicated that the vehicle can run smoothly along the curb line under a low control frequency in the greenhouse based on the method above. It was also found that this method could maintain better curb-following navigation accuracy even under interference of sundries if its length was not more than 300 mm, which may meet the actual need of vast majority of the production. In this paper, we proposed new technical ideas for robots running along curb with low cost in the greenhouse.